LED/Incandescent spectrum Watts/Photopic/Scotopic

I have had this information for awhile, in the form of power in "Watts" (color background is only for rough estimate/reference, colors background is not exact)



I then got to thinking a bit more, and decided to apply the human eye correction factors, to see how the eye would respond.

So I applied the human eye photopic function (daylight):



I then decided to apply the night adjusted human eye response (scotopic) to the same chart that you saw in the first one (which was "Watts"):



I spent a while looking back and forth between the charts, I thought it was interesting, I hope you do too.

if i understand well with the adapted night vision the amber led is like invisible at our eyes? :O

Humm, at least it appears the effect is rather minor with scotopic vision (night adapted eyes).

It looks to me like the LED's remain much more efficient for scotopic vision, because of the big blue spike... while the LED's simply look yellower than the incan, but cooler at the same time, due to the lack of red.

Sure enough, LED's go farther at night, and that's how I see the "warm white" LED Christmas strings at Costco -- yellower, but still cooler than clear incandescent strings.

IMO, if you are trying to simulate the look of incandescents with LED's and are restricted to standard white and one monochrome color, orange around 610nm might be a better bet.... only I think that might end up looking like high-pressure sodium, if that white/orange LED assembly I saw at Thornhill Broome Beach park (near Point Mugu, CA) is any indicator. Next time I'm around there, I have to go take a photo of it.

Canuke said:

It looks to me like the LED's remain much more efficient for scotopic vision, because of the big blue spike... while the LED's simply look yellower than the incan, but cooler at the same time, due to the lack of red.

Click to expand...

Well, by definition, 'scotopic' vision takes place in low-light conditions. The entire reason you've got the flashlight is so you can use your photopic visual system which needs good light to work. What the 'white' LEDs are good for is rapidly bleaching the rhodopsin from the retina and ruining your dark adaption. Yeah they've got a lot of light in the blue part of the spectrum, but it's way more light than the scotopic system can handle.

Since I seldom use a flashlight in dark conditions I prefer a red LED and low power incandescent flashlight as I can use these for brief periods without effecting my dark adaption.

FloggedSynapse said:

Well, by definition, 'scotopic' vision takes place in low-light conditions. The entire reason you've got the flashlight is so you can use your photopic visual system which needs good light to work.

Click to expand...

Not necessarily. If you use white/blue at extremely low levels, you can *use* instead of destroy night vision. I've been able to light up the entire forest with an Inova X1 that way, while camping, without losing night vision (so long as I don't hit something close by with it).

Of course, that's only good for seeing your way around generally... when you need to see something detailed (like a map or a dial) but otherwise wish to preserve night vision, that's where the deep red comes in handy.

Newbie, does a spectral graph of a "xenon" flashlight bulb look the same as the spectral graph of a 60W household bulb?

I have developed a ton of residential lighting sytems that use warm white LEDs. THis chain is very interesting because while the warm white LEDs can appear "warm" and match the color temperature of incandescent they are not rendering reds as well - their CRI or color rendering index is low. It is interesting that the discussion here discusses Scotopic vision and the concept of various colors of LEDs mixed together to make white light - if you look back into the early days of fluorescent light you will see that higher concentrations of violet and green caused problems with people with Scotopic Vision Syndrome. Companies trying to merge multi-color LEDs together to make true incandescent type lighting should be well aware of this effect. I made a light once that looked incredible under light meters in terms of color temperature (warm) and the color rendering (CRI) was very high - all the colors were rendered awesome...BUT - it caused certain people headaches...interesting! What I found was that certain people walk outside and put on sunglasses and are very affected by certain colors of light...what color tint do you like on your sunglasses?

ManuelLynch said:

I have developed a ton of residential lighting sytems that use warm white LEDs. THis chain is very interesting because while the warm white LEDs can appear "warm" and match the color temperature of incandescent they are not rendering reds as well - their CRI or color rendering index is low.

Click to expand...

Yes, that's a good point. Unless a special phosphor is used as is the case with the Luxeon warm white most warm white LEDs get their color temperature by just increasing the size of the yellow phosphor hump relative to the blue primary emitter hump. There's still a deficiency in the mid-range (greens) and still very little deep red produced. I find that deep red deficiencies are more tolerable in cool-white LEDs because by definition you expect cooler light sources to have less of their output in the reds. When this happens with warmer sources though it's positively headache inducing.

Canuke said:

IMO, if you are trying to simulate the look of incandescents with LED's and are restricted to standard white and one monochrome color, orange around 610nm might be a better bet.... only I think that might end up looking like high-pressure sodium, if that white/orange LED assembly I saw at Thornhill Broome Beach park (near Point Mugu, CA) is any indicator. Next time I'm around there, I have to go take a photo of it.

Click to expand...

It's hard to get incandescent-like light with cool white LEDs and one monochrome color. By the time you lower the color temperature enough it's very apparent that the light is mostly from monochrome LEDs.

In playing around trying to get HO train lighting for different era passenger cars (fluorescent for modern stuff, incandescent for pre-1960s) I've found that mixing warm white LEDs with 640 nm red in the proper proportion works nicely. The mix depends upon the relative outputs of the reds and warm whites. The effect more or less matches the 2700K to 2900K of standard incandescents. Those grain of wheat incandescent bulbs interestingly don't even work particularly well for simulating incandescent light. They're far too orange unless overdriven and they burn out in a couple of hours if overdriven enough to match household incandescent. I'm sooo glad to finally be rid of using those dreadful things in my model railroading hobby.

Not necessarily. If you use white/blue at extremely low levels, you can *use* instead of destroy night vision. I've been able to light up the entire forest with an Inova X1 that way, while camping, without losing night vision (so long as I don't hit something close by with it).

Of course, that's only good for seeing your way around generally... when you need to see something detailed (like a map or a dial) but otherwise wish to preserve night vision, that's where the deep red comes in handy.

Click to expand...

What I think would be interesting, though would be difficult to do well, would be to use a LED with two dies -- a blue die with phosphor "White" led, and a red LED, that could be activated separately, or together. Together, the red would serve to supplement the light from the white LED and provide better color-rendering. Alternatively, the red could be switch on only to allow map reading etc. without ruining night vision with a bright white hot spot.

OK, been meaning to comment on this. Can't sleep tonight so I might as well do so now. In addition to newbie's graphs I'd like to throw up a few more spectral curves for other common light sources. Before continuing with my thesis.

I ripped these graphs off from: gelighting

First off, sunlight. The yellow line in the graph


Next, old school incandescent



Finally, a triphoshor light:



Next are two links which detail some basic vision concepts. Rods cells, cone cells, sensitivity, etc.
http://en.wikipedia.org/wiki/Cone_cell
http://hyperphysics.phy-astr.gsu.edu/hbase/vision/rodcone.html#c1

Cone cells are used in good light. They are responsible for seeing color as well being sensitive to contrast and fine detail. When you use a flashlight it's because you want to engage the photopic visual system which needs good light to work. Understand that 98% of the cone cells in the eye are primarily sensitive to green and red light (they peak at 534 and 564 nm). The remaining 2% of the cone cells are sensitive to blue light (peak 420 nm). The blue cone cells are much more sensitive to light, but since there aren't nearly as many of them people cannot detect as much contrast in blue objects. This can be measured. Also, the blue cone cells are not concentrated in the center of the retina, like the red and green sensitive cone cells.This may be part of the reason 'blu-blocker' sunglasses help bring out contrast in objects (these glasses may also help reduce chromatic abberation too - blue light is not focused as well either).

Rod cells are used in poor light (scotopic, night vision). They are very sensitive to light, but have poor contrast and resolution compared to cone cells. Peak sensitivity is around 500 nm. Since night vision (in the lowest light levels) relies entirely on rod cells it's the reason you see no color (monochromatic).

The glare problem with white LEDs comes from the huge spike in the blue part of the spectrum, IMO. From about 440-460 nm there's a spike from the LED itself. A portion of this then hits the phosphor and creates the other (more useful) curve centered over the green/red part of the spectrum. So with white LEDs you've got two overlapping curves, one bright and tight in the blue part of the spectrum, and another longer shallower curve closer to the red part. It's this 'spike' in the blue part that really contributes to the glare of these lights, especially to dark adapted eyes. Radiation in the 440-460nm region is not very useful for contrast determination, and it's way too much light for the scotopic (dark adapted eye) to work with.

I think most people will agree there's a lot more to making a pleasing light than just its CRI number. I also think there's a lot to be said for 'continuous spectrum' lights. I think it's one reason people may prefer incandescents, even though the 'color balance' is not as good as flourescents. The incandesent bulb in some ways approximates daylight better as it's blue end is under control, and it provides more useful light in the green/red part of the spectrum. From a contrast point of view the wavelengths from about 480-500 nm through 620-650 nm are money, peak sensitivity. Too much blue just amounts to so much glare. The green part of the spectrum is weak in white LEDs. The sun peaks in the green part of the spectrum, so this may have an impact on the apparent contrast of these lights also.

OK, I'm rambling.. few more thoughts. Look at the graph for the last light, the triphosphor flourescent. While the output of this light would appear very 'white', because its whiteness is approximated by outputting in three narrow regions of the spectrum many people would describe the quality of the light as 'harsh' or 'artificial' even though it appears white. Human vision is very discriminating and can't be fooled that easily. The excess of blue light in many LED lights may also explain why they don't penetrate haze, fog and smoke as well - blue light is scattered more.

Anyway, interesting topic.

OK, time to continue my thesis. Nothing much new here today except a bunch of people yakking about brand A vs. brand B (same old, same old).

In my previous post I tried to explain why I think 'white' LEDs are so glaring and annoying to dark adapted eyes. It's primarily because of the blue spike in the spectrum, and the relatively weak greens.

Time for some more graphs. Here's the output of a 'full-spectrum' lamp from fullspectrumsolutions:



Of course, this is anything but 'full spectrum'. It's another multi-phosphor light. However it does provide a good approximation of sunlight to most people. Notice where most of the spectral energy is in this graph. Yes, it's in the green and red parts of the spectrum. Actually the three spikes in the graph roughly correspond to the peak sensitivity of the (blue, green, red) cones. Notice how the green and the red spikes are weighted much more heavily than the blue. Again, since over 95% of the cone cells in the eye are senstive to green and red (~530nm and ~575nm peaks) these radiation bands are most important for detecting contrast and for strain free reading. In a way this is taking the spectral output of a sun like source and 'sampling' it at the three regions corresponding to the three cone cells 'peaks' in the eye.

Below is a graph from venderbilt showing the spectrum of some common light sources (link).



The top graph is a blue LED + quantum dots, the second from top graph are the typical curves for a 'white' LED (blue LED + phosphor/scintillator). Notice how both of these have a 'notch' in the green part of the spectrum. Very bad, those wavelengths are important. If anything you want a peak in this part of the spectrum.

So, what to do? It's frustrating because in may ways white LEDs are closer to a true full spectrum source than just about anything out there. It just needs a little work. One obvious solution would be some sort of filter to attenuate the blue spike in the spectrum caused by the LED. The drawback would be slightly lower efficiency, and the need to tweak the phosphor to maintain the white balance. This might result in a more pleasing light. There's also the question of boosting the notch in the green part of the spectrum. Not certain how this could be done.

So it's tricky. It's possible to create a good approximation of white by mixing different colored LEDs... but the overall efficiency is not as good.

The efficiency of LEDs has finally surpassed most fluorescents (though perhaps not as good as high wattage incandescents - HPS and halide lights). The spectral output still needs tweaking though.

FloggedSynapse said:

It's frustrating because in may ways white LEDs are closer to a true full spectrum source than just about anything out there. It just needs a little work.

Click to expand...

I'm glad to finally see I'm not the only one to realize that. I see a lot of detractors saying LED light is poor, unnatural, etc. However, outside of the blue spike, slight dip in the green area, and lack of very deep reds it's way closer to sunlight than anything else at this point. Quantum dots seem very promising at overcoming most of the defects. Although I often 75 to 78 as the quoted CRI for LEDs I tend to think it varies with the color temperature. If the relative size of the yellow phosphor hump is increased to bring down the color temperature then the spectrum is much more balanced. I'd guestimate that 7000K LEDs might well have a CRI of around 75 but bins with a CCT closer to 5000K are probably in the 85 area. This is easily good enough for general lighting.

deleted

Slightly off topic sorry......but only slightly.

In the non flashlight electronics forum, I've had a thread about making a spectrometer from the sensor module (CIS) off a flatbed scanner. Hopefully I will be able to interface it to a 18F2550 PIC, with a USB interface and take some homemade spectrum readings. I've figured out 10 of the 12 pins on the sensor.....almost ready to make a cct for it.

This discussion got me thinking about more uses.....can the CRI and CIE coords be calculated automatically from a spectrum...if so, the what are the algorithms for doing this? Too complex for a PIC?

I'm hoping to build this thing into a general purpose lighting design test instrument....with a spectrum+CRI+CIE readings, what other sorts of readings would you think would be useful....for example:
- CRI calculated using photopic/scotopic...any other useful response curves
- accurate lux reading.......
- relative CRI (from a stored 'standard' spectrum), to see how much better/worse a test source is.
- some sort of triggering + time lapse function to show CIE drift with time/voltage etc

Incidentally...do the professional lab standard instruments show these data or do you have to download spectra into a PC program to do this sort of analysis....it'd be handy to have a handheld device, point it at a light and go....'this light has a CRI of xxxx'.

Phil

FloggedSynapse said:

OK, time to continue my thesis. Nothing much new here today except a bunch of people yakking about brand A vs. brand B (same old, same old).

In my previous post I tried to explain why I think 'white' LEDs are so glaring and annoying to dark adapted eyes. It's primarily because of the blue spike in the spectrum, and the relatively weak greens.

Time for some more graphs. Here's the output of a 'full-spectrum' lamp from fullspectrumsolutions:

Of course, this is anything but 'full spectrum'. It's another multi-phosphor light. However it does provide a good approximation of sunlight to most people. Notice where most of the spectral energy is in this graph. Yes, it's in the green and red parts of the spectrum. Actually the three spikes in the graph roughly correspond to the peak sensitivity of the (blue, green, red) cones. Notice how the green and the red spikes are weighted much more heavily than the blue. Again, since over 95% of the cone cells in the eye are senstive to green and red (~530nm and ~575nm peaks) these radiation bands are most important for detecting contrast and for strain free reading. In a way this is taking the spectral output of a sun like source and 'sampling' it at the three regions corresponding to the three cone cells 'peaks' in the eye.

Below is a graph from venderbilt showing the spectrum of some common light sources (link).

The top graph is a blue LED + quantum dots, the second from top graph are the typical curves for a 'white' LED (blue LED + phosphor/scintillator). Notice how both of these have a 'notch' in the green part of the spectrum. Very bad, those wavelengths are important. If anything you want a peak in this part of the spectrum.

So, what to do? It's frustrating because in may ways white LEDs are closer to a true full spectrum source than just about anything out there. It just needs a little work. One obvious solution would be some sort of filter to attenuate the blue spike in the spectrum caused by the LED. The drawback would be slightly lower efficiency, and the need to tweak the phosphor to maintain the white balance. This might result in a more pleasing light. There's also the question of boosting the notch in the green part of the spectrum. Not certain how this could be done.

So it's tricky. It's possible to create a good approximation of white by mixing different colored LEDs... but the overall efficiency is not as good.

The efficiency of LEDs has finally surpassed most fluorescents (though perhaps not as good as high wattage incandescents - HPS and halide lights). The spectral output still needs tweaking though.

Click to expand...

One has to really pay attention on those graphs. Example the first one is in uW, not eye corrected lumens or anything. Totally alters the whole thing once you fix it for human eye response...

NewBie said:

One has to really pay attention on those graphs. Example the first one is in uW, not eye corrected lumens or anything. Totally alters the whole thing once you fix it for human eye response...

Click to expand...

Hmmm.. I'd assumed all the graphs were absolute values, but I see you're correct and the GE ones are scaled to the eyes spectral response. Looks like I'll have to edit my thesis. :laughing:

The last graph from vanderbilt everything is scaled the same way. I was using sunlight as a model for an 'ideal' light source, so it was interesting to compare the spectral curves of the white LED to sunlight. It's a better match than incans or cfl's

I was using a dvd disk as a diffraction grating earlier today to look at the color spreads from various light sources. Comparing the cree LED to an incan filament was interesting. Actually got a pretty good spectrum off the cree. However the incan's spread extended much further off either end of the spectrum (there was little violet in the cree, but lots of violet in the incan. same thing for the incan reds, they extended much deeper). You could clearly see the green and red looked stronger in the incan, but the cree had a slight advantage in the blues.